United Nations Development Programme
Regional Bureau for Latin America and the Caribbean
Research for Public Policy
RPP LAC – MDGs and Poverty – 03/2008
Analysis on Relation between Natural Disaster Risks and Living
Conditions:
The Case of Bolivia
Ernesto Pérez de Rada*
Daniel Paz Fernández**
Octubre, 2008

Document prepared for the ISDR/RBLAC Research Project on Disaster Risk and Poverty. This
document is part of the Latin American contribution to the Global Assessment Report on Disaster
Risk Reduction, and the Regional Report on Disaster Risk and Poverty in Latin America. The terms
natural disaster and climate-related events will be used interchangeably, understanding that
socioeconomic conditions play a role to explain the intensity and consequences of such
phenomena. Thus, no event is strictly or exclusively natural.

UNDP-Bolivia ** Gobierno Municipal de La Paz, Bolivia.
The opinions expressed here are of the authors and not represent those of the RBLAC-UNDP.
Please cite this document as: Pérez de Rada, E. and D. Paz. 2008. “Analysis on Relation between
Natural Disaster Risks and Living Conditions: The Case of Bolivia”, RPP LAC – MDGs and Poverty
– 03/2008, RBLAC-UNDP, New York.
Working paper prepared for the UNDP-ISDR Project
Analysis on Relation Between Natural Disaster Risks and Living Conditions:
The Case of Bolivia1
FINAL DRAFT
Ernesto Pérez de Rada
Daniel Paz Fernández
October 2008
1
Working paper to be prepared for the “Disaster Risk Reduction Project” sponsored by UNDP and ISDR as part of the
Regional Studies on Poverty and Natural Disasters.
The authors acknowledge collaboration by Verónica Paz, Patricia Espinoza and Milenka Ocampo in elaboration of the
present working paper.
2
1. Background
The starting point of this document has been recognition of natural hazards importance and
their risk on development processes. Research faces an outlook that goes beyond improving
response capacities and designing mechanisms to deal with human aid or for emergency
assistance in case of disaster. UNDP (2004) has determined that disaster must necessarily
include study of impact on household living conditions, in order to thereby establish
improved mechanisms to mitigate and prevent its effects, as well as to sustain people living
conditions. Therefore, this is one of the main challenges in terms of learning, developing
tools and administration of development processes, particularly in regions particularly
exposed to natural hazards and their possible effect on human lives, on infrastructure and
income generation means of most vulnerable homes.
From this approach, several tasks are necessary in order to develop a frame for analysis and
management implied, both quantization of impact as well as aspects related to means of
recuperating productivity conditions and infrastructure affecting or at least endangering
living conditions of people exposed to them (De la Fuente, et. Al, 2008).
In the case of Bolivia, research on poverty and development conditions of households have
been mainly focused on elaboration of tools allowing to identify geographic location of
poverty – UBN2 and poverty line maps (UDAPE, 2002, 2005), national, department and
Municipal Human Development Indexes – as well as estimations of change in times of
incidence, the gap and intensity of poverty allowing to determine well-being evolution
patterns. That information has been used by several governmental administrations, which
in turn have developed strategies for poverty reduction allowing the elaboration of public
policy interventions in order to reduce poverty levels in the country.
From the point of view of estimation of determinants of well-being household conditions,
most of the poverty surveys have been oriented towards aspects of income generation, or
opportunities’ development through productive resources, as well as to growing capacities
such as education, health, understood as a human well-being requirements.
.
However, inclusion of weather hazards as determinants of poverty has not been studied in
the country. Estimates of natural events shocks over household, or mechanisms allowing
mitigating effects of climatic events, have not been part of surveys and research on poverty
in Bolivia. The only approaches developed have been the quantification of damages by
natural events such as inundation and drought, particularly during 2006 and 2007, when
large natural disasters occurred in the eastern and Amazonia region of the country (CEPAL,
2007).
The objective of this document is assessment of natural disaster impacts on income and
living conditions of exposed population, starting from available information from census
and household surveys. The present document stands at two levels of analysis; the first
involve the analysis of disaster risks at national and municipal level, while the second focus
2
Unmet Basic Needs
3
its attention in regions that have been affected by disaster in the eastern regions of the
country. In both cases it is intended to discover relations among natural events and changes
in living conditions. Given limited information, results may not be translated into causality
relations, and are simple estimates and benchmark to undertake future research containing
more thorough disaggregated information.
The document has been divided into five sections. Section 2 contains a methodological
background and information resources, while in section number 3, an approach has been
made on effects of natural disasters within the country’s municipalities. Section number 4
assesses household living in Trinidad city, the capital of Beni, which is the urban
conglomeration that underwent large disaster in 2007. Section number 5 underlines final
statements on study of Bolivia.
4
2. Methodology and information
Impact estimates of disaster on living conditions of the population can be approached from
two levels of analysis. On the first approach, main sources of information were taken from
population and housing3 census information, together with disaster information from the
DESINVENTAR4 data base and from other sources. Information contained on this first
part is referential and only takes into consideration municipal levels of disaggregation,
since disintegration of natural events only refers to this geographic level.
On second approach, natural events assessment is measured from of household surveys.
Nevertheless, limited information (see point 2.1) circumscribes the study to a particular
geographical area – the capital city of Beni – flooded between December 2006 and January
2007. Information on this particular case before and after the event is available. Since there
is no panel information, focus is made on the fact whether significant change is to be found
on well-being conditions before and after the natural event.
2.1. Available Information
The Bolivian case shows some particular characteristics, particularly in terms of available
information, which in turn prevent the use of several instruments for assessment of natural
event impact on household well-being. Disaster and natural and events, as well as
information on household surveys display characteristics that reduce precision of results as
well their validity.
Information on Disaster Risk
Three types of sources are available with regards to natural events. The first one is data
base made by the Civil Defence Vice-Ministry, in which natural phenomena are
systematized according to their type, department of occurrence and period. However,
problems deriving from this type of information are related to levels of disaggregation –
solely departmental – and temporary covering of data – that is, from 2000 to 2006 –
limiting global use of information with available well-being sources.
Second source is the DESIVENTAR data base, whose origin is written press news about
natural event registered from 1970 to 2007. Evidently this type of information is limited
since it is not supported by technical information - like registers from meteorological
institutions or geological research centres-. In fact, this type of available information is
strongly biased since there is an over reporting events of capital cities or regions with
secular events such as large flooding or dry periods. Additionally, DESINVENTAR data
base has a provincial scope. The mentioned geographical unit is in disuse at present, taking
into account the decentralization process effected since 1994, since which, official
3
The National Statistics Institute of Bolivia carried through two last Censuses in years 1992 and 2001.
The DESINVENTAR data base is a result of regional initiative taken by the Andean Information System for Disaster
Prevention and Attention.
4
5
information was systematized at municipal (corresponding to a section of a province) or at
departmental level (the totality of provinces).
On this account, an additional systematization work was carried through on
DESINVENTAR data base, with municipalities being identified according to event location
within a province. This implied loss of part of data with no precise identification of
municipalities involved.
Finally, there is information on climatic characteristics and probability of natural events on
the maps from the National Territorial Information System (SNIOT by its acronym in
Spanish) at the Ministry of Sustainable Development, which have been the basis for risk
geo referenced information systematising. To this effect, World Food Program (PMA by its
acronym in Spanish), enabled systematization of municipal occurrence probability of these
events starting from weather information, as well as agro-ecological SNIOT scale of
1:1,000. 000.
With this set of information sources, the strategy in order to estimate risks in Bolivia have
been as follows: it was determined that risks index need to be constructed from most
significant events found in available SNIOT information on flood, drought, and freezing
periods), assessed according to population exposed to them. Information contained at
DESINVENTAR data base is analyzed in terms of limitations and only as indicative data
on most important events identified by written press.
Information on Living Conditions
Living conditions of the population are available in two main sources of information:
population censuses, and household surveys. In the case of population and household
censuses, information has a nine year periodicity (1992 to 2001), and no information of
income is available. However, UDAPE simulation exercises combining survey information
on living conditions of the population (MECOVIs and EIH) and censuses provide income
imputations, allowing having municipal disaggregation information.
In the case of household surveys, information limitations are related to lack of
representativeness, which in turn implies the impossibility to infer results beyond urbanrural, capital cities and macro-region (highlands, valleys and tropics) levels. Likewise,
available information is a repeated cross-section, with no panel data, which implies the
presence of biased data when comparing data in time.
2.2 Non Causal Relations at Municipal Level
The first level of analysis, will focus on links between well-being indicators available from
official sources (see annex) and exposition to natural events. Study unit in this case is the
municipality and correlations will be esteemed from well-being changes, in order to obtain
non causal relations between natural disasters and changes in living conditions, according
to census data of 1992 and 2001.
6
UDAPE (2006) has calculated income imputations from household survey variable,
explained through information obtained in the censuses. Through this imputation poverty
lines for 1992 and 2001 years was obtained.
In the case of municipal risk assessment, risk events were obtained from SNIOT
information, and it were weighted by proportion of exposed population in municipality.
Hazard index is as follows:
Where risk index for i municipality (riesgo) is given by addition of the probability of k risk
(inundation, drought and freezing), normalized to 1, multiplied by index of population
exposed to risk. In the case of municipalities with a population below 50.000 inhabitants,
population index exposed is proportionate to municipality’s agricultural population,
whereas in the case of municipalities with over 50.000 inhabitants total population is
exposed to inundation (EXPOSITION=1), since usually whole municipality is exposed to
this type of hazard.
On this basis, municipalities with higher incidence of each risk have been geographically
referenced. Additionally, a pattern of municipalities was drawn with which to obtain a
categorization of changes in living conditions and environment.
For estimation of risk disaster effects on well-being, following function is proposed:
Where the depending variable is (inter-census) change on p, representing municipality i
well-being level. β represents well-being municipality risk indexes impact. γ show the
effect of control variables within municipality, susceptible to change in time (schooling,
migration, mortality, etc.). Finally, δ represents control effects on non-varying
characteristics inside municipality (altitude, slope of the terrain, productive vocation, etc.).
On account to possibly endogenous of well being risk variables, estimation strategies were
adopted starting from ordinary least squares, to simultaneous equations estimations (3SLS).
2.3. Impact of flood: the Case of Beni Department.
The second level of analysis of this document centres on estimation of changes on wellbeing indicators within a determined geographic area undergoing natural events of great
intensity. The unit of analysis is the city of Trinidad, capital of the Beni department. This
urban centre was chosen for two main reasons: first, Trinidad is the only recent case of a
capital city – for which representativeness of household survey data is acceptable - greatly
overflowed.
7
The second reason has to do with the fact that analysis of the event is currently important
since need for public policy assessment instruments for future execution stands out. In fact,
the Niño 2006-2007, generated impact with serious socio-economic policy implications in
the country, affecting over 100.000 families and provoking nearly 443 million dollars loss
(CEPAL, 2007). Characteristics of affectation in terms of territorial extension and intensity
in terms of damage and loss indicate such harm may hardly be exclusively attributed to
weather conditions in the country such as hydrological excess in the eastern region of the
country and deficit in occidental areas. Available information shows, on the contrary,
special relation between disaster and high degrees of vulnerability in physical, ecological,
social, institutional and economical factors in the country. This correlation concretizes,
therefore, in designing multiple risk and disaster scenarios particularly associated to
inundation and drought.
Damage in Trinidad city is directly associated with El Niño phenomenon, registered from
December 2006 to April 2007. Nonetheless, if state of emergency and disaster undergone
by the country during the same period the previous year is considered, it is evident that
effects of the latter are temporarily and territorially related to former disaster processes
magnifying and rendering risk conditions more complex. Additionally, current information
on meteorological and hydrological conditions foresees that extreme events in association
with hydrological excess and deficit in the country will gain greater intensity within
recurrent increasingly shorter periods.
Impact valuation strategy for the city of Trinidad was targeted on following exercises.
Strictly descriptive character analysis was achieved first, focusing on demonstrating
department socio-economic vulnerability, as well as relative change in living conditions
after flood in the city of Trinidad. The above implied estimation of change in monetary
poverty and extreme poverty, as well as intensity and severity.
On second term, proceedings went on to determine poverty patterns, according to multivarying regressions for both years. Estimations are as follows:
Where LnY is the head of family’s income logarithm depending on “district” residence
(distrito), whose own dichotomic scale of values are in accordance to district household
exposure to inundation. X represents set of control variables (sex, experience, education).
From these estimations for years 2006 and 2007, structural change in β regressors will be
tested to determine change in the way in which control variables at home and district levels
were affected according to poverty conditions that might be attributed to shock during the
8
two years. Magnitude and change in coefficient β is of particular interest, since it
represents a proxy of household exposition to risk. 5
Thirdly, inequality indicators for both years were compared. Mentioned indicators will be
chosen from generalized entropy family (GE). Absolute change on well-being conditions,
as well as vulnerability differences among certain strata and population groups facing
natural events, will be observed through exercise.
Finally, decomposition inequality on basis of regressions whose coefficients are used to
evaluate contributions relating to factorial inequality developed by Fields (1997) will be
calculated. These contributions are valid for several inequality measures, and various
explicative regressions variables (see annex for description of method employed). On this
particular case, Gini was used as inequality index.
Some authors also used methods based on regressions, but came short of number of
explicative variables they were able to use. Some only included one explicative variable
(Almeida and Barros (1991), Lam and Levison (1991)). Chiswick and Mincer (1972) only
included schooling, labour experience and number of working weeks and used log-variable
as inequality measure. On the other hand, Freeman (1980) used a large scale of variables
but decomposition is, according to his own words, “an approximation” and an “incomplete
decomposition”.
Therefore, the Fields method is particularly useful by the fact that it includes an unlimited
number of explicative variables and because relative contributions to factorial inequality
conditions thence deriving are valid for a large number of inequality measures. Particularly,
main objective in applying this method is determining contributing changes toward
inequality conditions deriving from shock inundation, whose proxy, will be the
neighbourhood district in the city. Aim pursued is that following contribution may be
statistically important, and that its relative importance in explaining inequality might be
incremented along inter-annual period under study.
5
Lack of panel data or even a pseudo-panel construction – on account of reduced sample - two years – implies bias
deriving from assessment of weather effects. Nevertheless, magnitude of event registered in Trinidad, is presumably large
enough as to provide an idea of effect on living conditions.
9
3.
Analysis at Municipal Level
3.1 Analysis on Risk Incidence in Bolivia
Disaster Data Systematization
DESINVENTAR data from an inter-census period (1992-2002), was systematized in order
to assess an event’s incidence on population well-being. There are well-being indicators6
(UBN, poverty line, IDH), that may be related to events occurred.
Still, available information is limited. From over more than 1.600 DESINVENTAR events
registered between 1970 and 2007, only 289 belong to the 1992-2001 inter-census period,
while solely 281 are relevant. In 84 from the 389 cases registered information does not
correspond to one natural disaster event, while from remaining sample (305), 24 had no
precise location. On the other hand, 158 from 314 municipalities register at least 1 from
281 events, which means that sample only represents 50,6 % of municipalities. Most
extreme case belongs to Pando, where only one event was registered in the capital city of
Cobija from 1992 to 2001, while it is generally known that this department undergoes
yearly intensive rainfall and inundation.
On this account DESINVENTAR does not permit precise identification of municipalities
demonstrating greater degree of hazard and natural disaster vulnerability, neither does it
allow translating tendencies, on account of disproportionate bias of events registered in city
capitals (142) and the rest (139). Over half the events took place in 9 municipalities
(2,6%), La Paz being the one with highest registers (62).
This is, fundamentally, because DESINVENTAR data base identifies events only from one
source, La Paz morning paper “El Diario”, which, while being a country wide circulating
newspaper, reflects most of its information on events happening in that city. Also, very
often there is distorted information, on account of description of event’s impact not often
being assessable. This is often the case in rural zones, where (frequently due to absence or
inadequate use or loss of information) direct impact on agriculture, forestry, education,
health, etc. may not be assessed.
DESINVENTAR limitations
DESINVENTAR information was systematized into thematic maps showing the number of
events per municipality for each type of disaster.7
6
7
On basis of official INE and UDAPE information.
See Maps on next page.
10
Following Maps Nos. 1, 2 and.3, were drawn on basis of DESINVENTAR
Municipality Data Systematization corresponding to Inter-Census Period of 19922001.
Map No. 1. Floods
Map. No. 2. Droughts
Map No. 3 Events of largest impact
Map No. 1 displays the number of Inundations Registered in Different Bolivian
Municipalities As shown, most vulnerable regions in terms of inundation are located
within the Amazonas territory (Beni, Santa Cruz and La Paz departments) as well as in that
of Tarija. However, map information does not correspond to reality and there are some
elements which might cause confusion and incorrect interpretation. Municipality of La
Paz, (registering 23 inundation episodes, appears as the municipality in darker blue, while
municipalities in the department of Beni (registering one only inundation event each one)
appears in light blue. On a quick sight one could interpret that La Paz is by far the
municipality with greater number of inundations, however, this is due to disproportionate
biased available information between municipalities. Surprisingly, the department of
Pando (north of the country) appears in white, meaning that none of its municipalities
registered any inundation during that period.
Map No.2, shows Number of Dry Periods Registered in Different Bolivian
Municipalities. Although information in this case may seem to be consistent, drought
seem to affect only the oriental part of the country, corresponding to el Chaco and
subtropical valleys of Santa Cruz and Tarija. The department of Potosí, on the contrary,
presents only one municipality registering drought, while the Altiplano is one of the driest
regions not only in the country, but in the whole continent. 80% of Potosí territory is
located in the Altiplano and some of the municipalities in that zone have less than 15 days
of rainfall a year.
Map No, 3, shows Greatest Impact Event in Different Bolivian Municipalities. In this
case, inundations appear in blue, dry periods in yellow and freezing periods in light pink.
Quick reading of this map permits visualizing geographic macro regions affected by
different natural disasters. South-western municipalities in the country, corresponding
essentially to the department of Potosí, show freezing impact; municipalities located to the
11
north of the country, show impact due to inundation, while municipalities situated to the
east, display impact on account of drought.
Still, reading is relative since, for instance, in the case of Potosí, although drought seems to
be the event of greatest impact, lack of information on droughts could make it be
mistakenly interpreted, in that Potosí is a department with exclusively “freezing” periods.
However, given DESINVENTAR characteristics in the case of Bolivia, it is advisable to
only keep register of mentioned events as chronological data, not to be used as tendency
reference of municipality events towards future local public policy planning.
By above exposed, DESINVENTAR is unreliable for estimation of natural disaster
behaviour tendencies in Bolivian municipalities. Information is incomplete, and
interpretation of different results obtained through this source could generate confusion and
lead to mistaken interpretation.
On Non Causal Relations at Different Geographic Scales
Under this context, information taken into account to determine municipality tendencies in
terms of risk and natural disasters, will be one elaborated by World Food Program (PMA
by its acronym in Spanish), based on systematization of agro-ecological maps from the
Sistema Nacional de Informacion de Ordenamiento Territorial (SNIOT) elaborated by the
Unidad de Ordenamiento Territorial from the former Ministry of Sustainable Development
in year 1997.
In this case, the PMA assessed yearly probability of existing natural risk. For instance,
municipality of San Pedro de Quemes located to the south of Potosí and on the border with
Chile (Andean Cordillera), with a freezing frequency of 270 days a year, permits creating a
freezing risk index ponderable starting from interaction between (freezing) risk probability
and municipality endangered population. Still, it must be said that this does not allow for
evaluation of damage to infrastructure or agricultural plantations.
Once index 1) freezing risk, has been assessed it is proceeded to evaluate indexes 2)
drought risk and 3) inundation risk. From calculations, assessment of, municipal
vulnerability to general risk may be determined, by adding up values found in three
indexes.
It is very important to point out that natural disaster events may be of different types and
nature. Still, on account of source unreliability and scarce information, consideration of
this paper will be three types of natural disaster events most recurrent in time and space,
from which behaviour tendencies and risk indexes may be elaborated. These three disaster
types are freezing periods, drought and inundation.
Behaviour and Tendency of Natural Disaster
12
Maps Nos. 4, 5 and 6 displayed on next page, were based on PMA Systematization
Data and on Different Municipality Risk Indexes Assessed.
Map No. 4 Freezing Risk
Map No. 5 Drought Risk
Map No. 6 Inundation risk
Map No. 4, shows Freezing Risk Present in Different Bolivian Municipalities. As may
be seen, freezing affects two distinct regions in Bolivia. First one is Highlands comprising
south of the La Paz department, while the other two are Oruro and Potosí, the second one
being Valles Altos (High Valleys) which include most of the departments of Chuquisaca,
Tarija and Cochabamba. Incidence of freezing risk strongly depends on altitude above sea
level at which different municipalities struck are located. Thereby, definite “East-West”
echeloning may be appreciated in relation to territory freezing risk. Municipalities
appearing in white are located at an altitude below 2400 meters above sea level and do not
run any risk. Freezing risk is increased at higher altitudes, and in municipalities on the
border of Chile, located between 4000 and 6000 meters above sea level, risk becomes
critical (dark purple).
Map No. 5, shows Drought Risk Present in Different Bolivian Municipalities. In this
case it may be evidenced that drought mainly affects the south of territory and involves
regions of Altiplano (highlands), Valleys (High and Low) and El Chaco. The Altiplano
region characterizes for being arid due to scarce oxygen on account of altitude. The
Valleys (High and Low) though they may be humid zones at some seasons of the year,
during part of autumn and winter, are severely struck by drought, with negative effects on
agriculture and cattle-raising.. Finally the Chaco region, located to the south of Santa Cruz
department and to the east of Tarija, undergoes severe drought during most part of the year.
Some Chaco regions have below 15 days of rainfall a year.
Map No. 6, shows Risk of Inundation Present in Different Bolivian Municipalities.
Although risk is valid for all Bolivian departments, it is clear that Amazone departments of
Pando, Beni, north of La Paz and Santa Cruz are most affected by this type of disaster.
Absolutely all municipalities in Beni and Pando display inundation risk; and at some of
these municipalities disaster, could be of considerable magnitude, since response capacities
are almost non existent at some very isolated communities.
13
As may be observed from these three maps, almost all Bolivian municipalities are prone to
some type of risk. Amazone municipalities are most affected by inundation, while
municipalities to the south and those of the Altiplano are affected by drought, and
municipalities in the Altiplano alone are mainly affected by freezing.
From former evaluation, vulnerability to natural risk disasters behaviour and tendencies
may be assessed. Next, municipalities most vulnerable to overall risk will be assessed,
according to freezing, drought and inundation periods normalized addition.
Map No. 7 General Index of Risk by Municipality
Final result displays that municipalities at the highlands, chiefly those located on the
Chilean border, the interior of the Oruro and Potosí departments, are most vulnerable to
natural risk from the whole territory. Likewise, that the south and centre of the country,
where departments of Tarija and Sucre are located, has a somewhat high index of
vulnerability. Vulnerability index is moderate within Beni department, while municipalities
in La Paz, Cochabamba, Pando and Santa Cruz, mostly display low degrees of
vulnerability.
14
Though results of these tendencies are real, risk impact is not the same on the whole
territory. For instance, departments of Oruro and Potosí (appearing in darker red) are most
vulnerable to all types of risk (there is a very high risk of drought and freezing within their
municipalities, together with a moderate risk of inundation). However, more often,
population exposed to those types of disasters is reduced, and economic activities are rare.
Consequently, while these municipalities display high risk indexes, conclusions are relative
since real impact on human activities may come to be almost none in least populated
municipalities.
On the other hand, there are Beni and Pando municipalities displaying low and moderate
risk, where disaster impact is nonetheless much higher than in other municipalities of the
country. In such a case, municipalities at the Amazone region do not display freezing, or
drought, but where magnitude of inundation is such that it sometimes reaches enormous
proportions. During 2006 and 2007 rains in Beni department, over 15000 people were
directly affected by El Niño, and economic loss in agricultural and cattle-raising areas, was
of millions.
Assessment of natural disaster in Bolivia tendencies may be concluded by saying that over
70% municipalities in the country are exposed to some type of event. Main types of
disaster having an effect on population are freezing, drought and inundation. The Altiplano
is the most vulnerable region in Bolivia, while it not being most impacted. Amazone
region, in change, while not being the most vulnerable one, is the one undergoing largest
impact.
On Cross Sectioning Disaster and Well-Being Indicators
Once natural disaster incidence overall behaviour has been established in Bolivian
municipalities, attempts have been made to identify whether there is direct relation between
disaster incidence and well-being of population exposed to disaster. It is simply a matter of
identifying existing correlation between disaster and living conditions of damaged
population
In order to achieve this task, data on risk index (freezing, drought, inundation) from a
determined municipality was cross-sectioned against population well-being indicators, such
as IDH, Poverty Line or the NBI, registered in the inter-census period (1992-2001), by first
considering overall conditions of the population on initial period (1992) to subsequently
assess overall conditions at the end of inter-census period (2001).
Results from cross information shall display municipal general tendencies on people living
conditions following disaster. Tendencies will display whether improvement of population
living conditions followed after disaster, and if living conditions and habitat were
deteriorated or not within affected municipalities.
Following graphic shows overall evolution tendency of living conditions in municipalities
facing natural disaster. General parameters used to achieve this tendency are 1) Disaster
Risk, and 2) Change in Poverty Line following disaster.
15
.4
.2
0
-.2
-.4
0
.2
.4
.6
.8
1
riesgo1n
Graph. No. 1. General patern between change in poverty and risk index
Though it may seem there is a tendency indicating that to higher risk index, living
conditions will be worsened, it is not always the case.
Synthetically reading this graphic, may result in classification of municipalities by types,
according to evolution of living conditions opposed to natural disaster. To this effect, 4
types of municipalities may be identified displaying different degrees of evolution (of their
living conditions) with regards to existing disaster risk.
Previously 4 mentioned municipality types are:
.
High Risk Index Municipalities with Improved Living Conditions (Dark Blue)
This is the group with least number of municipalities. Exceptional though this may
seem with respect to general tendency, this may happen.
.
Low Risk Index Municipalities and Improved Living Conditions (Light Blue)
This group responds to a logic. The least exposed to risk a municipality is, the more
chance will it’s population have to improve its living conditions.
.
High Risk Index Municipalities and Worsened Living Conditions (Light Red)
16
This group of municipalities responds to a different logic. The more exposed
municipality is to risk, the fewer chances will it’s population have of improving living
conditions.
.
Low Risk Index Municipalities and Worsened Living Conditions ( Dark Red)
These Municipalities also escape to general tendencies. In this case, they are lowest
income municipalities and the most abandoned in the territory.
Map No. 8 Tipology of Municipalities under poverty change and risk criteria
As the Map shows, Municipalities which though having a high risk index improve their
living conditions are to be found in La Paz, Oruro Cochabamba and Chuquisaca.
This may be explained by the fact that impact of event (in this case drought or freezing), is
not as harsh in terms of population or produce affected.
Municipalities displaying low risk index and improvement of living conditions are chiefly
located in Tarija, Chuquisaca, Cochabamba, Santa Cruz, Beni and Pando. In some cases
this is on account of impact of event reaching very small towns.
17
Municipalities showing a high risk index with worsened living conditions are mainly
located in the Altiplano. The reason for this is that risk indexes in these municipalities are
extremely high.
Finally municipalities, that while not showing great vulnerability to risks, worsen their
living conditions, are mainly located to the north of La Paz, Beni, Pando and Santa Cruz. It
is the case of municipalities having dispersed population scarcely communicated, which
explains reaction to disaster being slow, a fact worsening well-being conditions.
Furthermore, there are other factors weighing on well-being conditions, such as migration
rates, productivity possibilities and access to markets in different capital cities belonging to
a municipality.
3.2 Living Conditions and Natural Invents Change Pattern
In order to design a pattern for natural event impact, change in municipal poverty incidence
was assessed according to several specifications, including natural risk variable. Reduced
scale expresses that:
(3.1)
Where Pt-Pt-1 is change in poverty incidence (FTG(0)) in municipality i, depending on the
following variables:






P92, is municipal poverty incidence on initial year (1992).
NBI Change, which is change of municipal unmet basic need indexes (non monetary)
during inter-census period.
TMN, which is municipality migration rate during inter-census period.
Cambioeduca, is change of municipality average school years ended by population
above 19 years of age.
Risk, being municipality risk rate combined with drought, inundation and freezing,
considered at the above section.
Mortality Change, is change in municipal child mortality rates.
There are some difficulties in estimation of equation 3.1. There may possibly be
endogenous biases among some variables. There are particularly two regressors that may
show this characteristic. First one is risk rate, since in model explained, exogenous quality
is assumed with relation to poverty incidence. Nonetheless, according to conceptual frame
developed by UNDP (2004), it is established that risks are reinforced by poverty, since the
poorer population is, the more it will be exposed to natural risks. Therefore risks and
poverty are increasing factors when they go together.
18
Other variable that may be endogenous is NBI. According to the model, it is presumed that
improve on non monetary living conditions should lead to improved monetary poverty
conditions. However, improvement in monetary conditions may also be a factor affecting
non monetary well-being (especially that related to housing conditions and access to health
services).
The following strategy was adopted in the attempt to correct these problems. On a first
design, model was calculated from Ordinary Least Squares (MCO in Spanish) in order to
adopt a benchmark model. On second stage, an estimation of Seemingly Unrelated
Regression model was adopted (SURE) where system includes equations in which NBI
variables on risk and change in UBN are also modelled. Third estimation used SURE
corrected by degrees of freedom. (SLS) Two Stage Least Squares method was applied to
fourth estimation based only on poverty change and risk. Finally at fifth estimation,
simultaneous equations method (3SLS) was used.
The rest of equations of the model were proposed as follows:
(3.2)
Equation 3.2 has UBN change (NBI) as dependent variable, which depends on:


Risk index (riesgo) and
Poverty change (cambiopobreza)
Finally risk equation is as follows:
(3.3)
Equation No. 3.3 models risk as dependent variable as a function of exogenous variables
related to climatic and geological aspects within the municipality.



Altitude (altura)
rain precipitation (precipitacion) and
municipality average area with over 30 degrees gradient (pendiente).
In the case of equation 3.2, it is assumed that non-monetary well-being conditions are also
related to change in monetary poverty and risk, while in equation 3.3 risk index is
calculated through exogenous variables (non-varying in time) such as altitude, precipitation
and gradient.
Results of these estimations are displayed on chart 3.1.
19
CHART 3.1.
ESTIMATIONS OF CHANGE IN POVERTY INCIDENCE AT MUNICIPALITY
LEVEL
MCO
SURE
SURE - AJUSTE
GRAD. LIBER.
2SLS
SIM. EQ (3SLS)
Equation 1:
P92
.0910804
(.0514194)
.0219261
(.0026578)
.0002364
(.0003314)
.0360562
(.0120107)
-.1169673
(.0189808)
-.0004538
(.0002466)
-.0551276
(.0400289)
0.3107
22.84
311
.059492
(.0459805)
.0366192
(.002413)
.0001365
(.0002962)
.0260058
(.0107553)
-.1106567
(.0181587)
-.0003248
(.0002206)
-.0215602
(.036121)
0.2349
303.18
311
.0593782
(.0465069)
.0367365
(.0024406)
.0001361
(.0002996)
.0259728
(.0108785)
-.1106353
(.0183666)
-.0003244
(.0002231)
-.0214425
(.0365345)
0.2337
49.65
311
.1227646
(.1136482)
.0727613
(.0166978)
-.0002555
(.0210075)
.0024577
(.0005596)
-.0415988
(.00529006)
-.0012642
(.0004555)
.0913149
(.0784839)
0.3459
6.19
311
.0200725
(.0309786)
.0774851
(.010643)
-.0000197
(.000133)
.0003079
(.0048315)
-.030789
(.00320285)
-.0001284
(.0002017 )
.0222874
(.0278203)
0.2133
333.02
311
Riesgo
-
-.9791202
(.316987)
14.02038
(.91753)
-.4552053
(.2221089)
0.1118
117.45
311
-
Cambio
pobreza
Constante
-.9827519
(.3154544)
14.0645
(.9130939 )
-.4569783
(.2210351)
0.1105
238.67
311
-
-.3830289
(.442533)
12.7839
(1.948239)
-.1014442
(.2897237)
0.1274
44.83
311
.0000875
(.0000195)
-.0018668
(.0005188)
-.0026111
(.0005507)
.654643
(.0869502)
0.3275
154.38
311
.0000875
(.0000197)
-.0018668
(.0005222)
-.0026112
(.0005542)
.654641
(.0875148)
0.3275
50.80
311
.0000854
(.0000197)
-.0019011
(.0005241)
-.0025139
(.0005563)
.6589137
(.0878268)
0.3276
49.85
311
.0000847
(.0000195)
-.0019263
(.0005173)
-.0026382
(.0005478)
.6675938
(.0868109)
0.3275
153.68
311
Cambio-NBI
TNM
Cambio-educa
Riesgo
Cambio-Mort.
Constante
R2
F/Chi2
Obs.
Equation 2:
R2
F/Chi2
Obs.
Equation 3:
Altura
-
Precipitación
-
Pendiente
-
Constante
-
R2
F/Chi2
Obs.
-
-
Note: values in parenthesis stand for standard errors.
Without assuming causal relations, given limited information, grade of aggregation and
temporary timing of analysis, in a general way, it is evident that in equation 1, different
methods of estimation reveal that variables establish expected results. In terms of statistical
significance, both risk variable as well as schooling change are statistically significant in
any of the models. From regressions, it is clear that:

Schooling level shows increasing contribution to improve poverty incidence.
20





The fact of an initially higher municipal poverty incidence has a positive effect on
change of same with time. However this result may only be interpreted as an effect
of been initially situated at a disadvantageous position, and not as an index of
municipality convergence.
Negative change in infantile mortality rate (interpreted as improvement) together
with positive migration rates (immigration) positively influence poverty incidence.
However direction of sign at both variables changes with simultaneous calculation
and two stages estimation, though significance also falls.
In the case NBI change, improvement of municipal poverty conditions has
positively relation to NBI change, significance remaining unaltered in any of two
models.
Finally risk variable, shows expected results from theory, meaning that, to greater
municipality risk, correspond lesser improvement of living and monetary
conditions. Variable shows negative and statistically significant values upon five
estimations, though in the case of two stage method, and in that of simultaneous
equations, magnitude of the effect measured by the coefficient, falls in more than
50%, situating itself between 4% and 3% respectively.
Lastly, it must be said that results of equations 3.2. and 3.3. are qualitatively
consistent in their different estimations, emphasizing the fact that risk index has a
negative incidence on NBI improvement.
21
4.
WELL-BEING CHANGES IN TRINIDAD AFTER THE EL NIÑO 2006-2007
PHENOMENON
4.1 Socio Economic Context in Beni
Department of Beni located to the north of the Republic of Bolivia, has an extension of
231.264 Km2. With a projected population of 422.434 inhabitants for year 2007, it
concentrates 4.3% of national population. Beni department is an “expulsing” department,
with a negative migration rate of 8,4 for each thousand inhabitants. Political distribution
divides the department into 8 provinces and 19 municipality sections.
Map No. 9 Beni: Political Division
Demographic and economic profile of the department is characterized by a young, urban
agricultural, cattle raising population with high migration rates. High economic dependency
on agricultural and cattle raising activities makes economy in Beni highly vulnerable to
climatic risks, mainly, inundations and drought. To this characteristic, population
composition mainly belonging to younger generations must be added, a factor accelerating
impact of natural disaster.
Natural disaster impact in this department, specially that derived form the El Niño
phenomenon in 2006-2007 reveals necessity of orienting public policy toward actions of
risk prevention, as effective instruments for poverty reduction. Additionally risk and
natural disaster prevention might constitute an effective instrument for growth of
economic capacity by reducing impact of infrastructure and basic services loss –
particularly water supply and drainage systems – and consequent serious diseases.
22
An Agricultural and Cattle-Raising Economy
Economy in Beni is based on agricultural and cattle-raising activities. Contribution of these
activities represented 38,5% of the departmental produce for year 2007. During the last
two decades this sector was first contributing to departmental GDP; in year 1988 this
sector’s contribution already represented 37% of produce. Second economic sector in Beni
is manufacturing industry contributing with 17% to department PIB. Following activities
are commerce with 11%, and public administration with 10%. Economic dependence of
agricultural and cattle-raising sector transcends the field of produce generation: 18
Agricultural and Cattle Raising Peasant Economic Organizations (OECAS) are registered
in the rural areas in agriculture and cattle-raising, handicrafts, and forestry activities.
Additionally agriculture and cattle-raising constitute the chief source of income from
exports in the department; chestnut exports alone represent more than 50% of department
exports.
Graphic No.4.1
Contribution to GDP According to Economic Activity (%)
Department contribution to national product in 19 years oscillated between 4,2% and
(1988) and 3,5% (2007). Along this period, economic activity in Beni kept development
rates below national averages. Under high dependency circumstances on agricultural and
cattle-breeding and raising, natural disaster affecting Beni inhabitants year by year, may
constitute, on the short term, in important drawbacks for department development.
An example of such vulnerability is impact registered by the El Niño phenomenon 20062007. Loss of cattle raising sector was quantified in US$ 31 million dollars, owing to loss
of 178.000 livestock from 3 million registered in Livestock Cadastre in year 2006.
23
Productivity vocation of the department shows large diversification potential of economic
activities towards activities transcending economic insertion in merely extractive activities.
Sectors that may be mentioned for their great potentiality to generate department surplus,
are forestry exploitation other than for wooden purposes (e.g. chestnut), expansion of
organic goods production and tourism infrastructure. Target of political policies in the
department should be supporting these activities which combine sustainable environmental
and labour strategies with strategies for risk and disaster prevention.
Young population
Beni population is characterized by a young demographic profile. Population composition
shows that it mainly concentrates in the range below 5 years, corresponding to 15% of
department population. Previously mentioned is important in terms of population
vulnerable to natural disaster impact. According to several surveys on impact of natural
disaster, infantile population results in one of the groups particularly vulnerable to
environmental impact. There are several factors determining identification of this group as
part of the population particularly vulnerable: death during disaster, disease proliferation
and larger malnutrition risk, and death of people at their charge (UNDP, 2007; UNDP,
2004).
Contributing to this young age composition characteristic, delay in compliance of ODM
related to development and living conditions of population may be added. Chronic
malnutrition and maternal mortality are above national averages. Last data registered on
2003 reported 28% chronic malnutrition, 4% above national average; something similar
happens with registers on maternal mortality which reported the third highest index for year
2005, with 259 deaths for every 100,000 live newborn, compared to national average of
234. Despite these indicators, department infantile mortality stays below national average.
Important data in terms of effect of natural disaster is, no doubt, high incidence of serious
diseases: malaria, which despite of having favourably diminished since 2001, still highly
affects Beni inhabitants, and is seven times above national levels. Propagation of serious
disease in disaster seasons is a fact; according to the Malaria Prevention National Program,
levels of infestation grew during the last months of 2006 and first of 2007, as a
consequence of the El Niño Phenomenon. Dengue Propagation is another example of risk
in disaster seasons.
Basic Sanitation and Drinking Water Access Deficit
Living conditions of Beni population is typified by the lowest coverage rates with regards
to sanitation system and potable water, together with the department of Pando. Compliance
with seventh ODM for year 2005, situated Beni in the penultimate place in drinking water
coverage and at the last place in draining system coverage. Only 45% of population has
access to drinking water and only 15% has access to sanitation. Access opportunities to
both services in Beni department are far below national average.
24
Graphic 4.2
Potable Water Coverage Rate (%)
Meta Nacional al
2015: 78,5%
71,7
57,5
BOLIVIA
83,4
SANTA CRUZ
69,1
81,9
LA PAZ
57,9
76,3
TARIJA
60,2
73,0
ORURO
63,2
62,8
CHUQUISACA
40,7
62,7
POTOSÍ
40,1
52,5
COCHABAMBA
44,0
45,4
BENI
33,0
36,2
PANDO
24,0
0
10
20
30
40
50
60
70
80
90
%
1992
2005
Graphic 4.3
Sanitation Coverage (%)
Meta Nacional
al 2015: 64,0%
43,5
28,0
BOLIVIA
60,8
LA PAZ
32,3
53,7
TARIJA
39,1
CHUQUISACA
41,0
28,4
40,6
COCHABAMBA
32,8
33,6
SANTA CRUZ
24,2
ORURO
33,5
17,7
32,7
POTOSÍ
19,9
29,6
PANDO
25,5
28,2
BENI
15,2
0
10
20
30
40
%
50
60
1992
70
2005
It is not only that there is shortage in potable water provision, but its distribution also
displays important inequality rates. From 19 municipalities in Beni, only 5, that is between
52% to 72%, have coverage, 3, San Borja, San Ignacio and San Ramón) have between 34%
to 52% coverage, while the rest is unable to cover 34% of municipal population. Between
1992 and 2001, potable water coverage rose in all departments, however there are persistent
uneven regional opportunities of access to this service: Santa Cruz and La Paz have
25
coverage surpassing 80%, while in Beni and Pando departments it only reaches 50% of the
population. These differences could be attributed to population low demographic density
and dispersion in two departments with lowest coverage.
During year 2005 Beni department had a sanitation service coverage of 28,2%, having
positively evolved since 1992. Between 1992 and 2005, department basic sanitation
coverage only grew in 13%, being below last national average observed (43%) and almost
36% from national goal (64%).
Natural Disaster and Contingency Plan
Most important regional impact from climatic drought and inundations phenomenon occurs
in Beni. With the objective of creating response capacities to emergencies, through
collaboration from the Vice Ministry of Civil Defence and Integral Development, together
with World Food Program (PMA by its acronym in Spanish) have developed the national
contingency plan as well as department contingency plans in order to identify hypothetic
scenarios of expected disaster in next five years.
Joint effort by national and departmental planning offices has developed a method for
qualification of alimentary crisis, by taking into account resources and response capacities
to emergency situations in the Beni department. Plan’s main objective is providing
alimentary assistance during climatic event and at recuperating stage; and safeguarding
nutritional level of most vulnerable groups.
Chart 4.1
Drought Probability and Impact
Provincia
Municipality
Magdalena
Iténez
Baures
Huacareja
San Joaquín
Mamoré
San Ramón
Puerto Siles
Trinidad
Cercado
San Javier
San Andrés
Marbán
Loreto
San Ignacio e Moxos
Moxos
Territorio Indígena Parque
Nacional Isiboro-Sécure
Santa Ana de Yacuma
Yacuma
Exaltación
San Borja
Reyes
José Ballivián
Rurrenabaque
Santa Rosa
Riberalta
Vaca Diéz
Guayaramerín
Source: PMA (2008)
1: improbable; 2: probable; 3: very probable
B: large impact, C: moderate impact
Probability
3
3
3
3
3
3
3
3
3
3
3
Impact
B
B
B
B
B
B
B
B
B
B
B
1
C
3
2
3
3
2
3
3
2
B
B
B
B
C
B
B
C
26
Impact on 98,183 (19,637 families) has been foreseen in case of drought. Distribution of
food will reach 9,428 metric tons. In the case of Beni, beneficiaries of the contingency plan
for inundation amount to 31,428 people (6,286 families). 1,520 metric tons of food will be
delivered during recovery.
Previous table shows probability of occurrence and drought impact in Beni provinces and
municipalities. Scale employed reveals high probability of drought occurrence, as well as
greater impact, in terms of density. It is evident that vulnerability scenario affects all
municipalities in this department. To this high probability of drought occurrence,
inundation probability menacing department municipalities may be added, though with
increasingly varied occurrence and impact intensity.
Chart 4.2.
Inundation Probability and Impact
River
Mamoré e Ichilo
Province
Marban
Moxos
Cercado
Mamoré
Vaca Diez
Iténez
Beni
José Ballivián
Iténez
Vaca Diez
Iténez
Mamoré
Municipality
San Andrés
Loreto
San Ignacio de Moxos
Trinidad
San Javier
Puerto Siles
San Joaquín
San Ramón
Santa Ana de Yacuma
Exaltación
Guayaramerín
Baures
Huacaraje
Magdalena
Rurrenabaque
Reyes
Santa Rosa
San Borja
Riberalta
Baures
Huacaraje
Magdalena
San Ramón
San Joaquín
Baures
Guayaramerín
Vaca Diez
Source: PMA (2008)
1: improbable; 2: probable; 3: very probable
B: large impact, C: moderate impact
Probability
3
3
3
3
3
3
3
3
3
3
3
3
3
3
2
2
2
2
2
2
2
2
2
2
2
2
Impact
A
B
A
A
B
A
B
B
A
A
C
B
C
B
A
B
B
A
B
A
B
A
B
B
C
C
Department regions more prone to disaster correspond to productive spaces found on the
banks of Mamoré, Iténez and Beni rivers. Worst prognosticated scenery in simulations is
the one corresponding to the Mamoré delta. Main consequences of inundation is worsened
availability and access to food products with consequent loss of income source: in this
setting most vulnerable groups are inhabitants whose income and/or jobs depend on
agricultural, cattle and fish breeding.
Contingency plan for the department foresees that 31.428 people from 6.286 families
benefit. On first stage, 1.031 metric tons of food will be distributed along 60 days: On
second stage, foreseen for a period of 120 days, 488 metric tons will be delivered.
27
Above table displays probability of inundation occurrence and impact in provinces and
municipalities, mainly in basins of Ichilo and Mamoré rivers. Impact intensity is varying.
4.2. Estimation of Impact of Inundation on Trinidad
Data used in the analysis comes from 2006-2007 Survey on Improvement of Living
Conditions (MECOVI, by its acronym in Spanish). These surveys were applied by the
National Statistics Institute in large Bolivian cities, including Trinidad. As mentioned in
the methodological section of this work, survey has the necessary statistical representation
with regards to this city.
For purposes of this paper, household labour as well as non-labour income was selected
from surveys. Size of sample varies from 373 observations in 2006 to 333 individuals in
2007.
Descriptive Statistics
Population. In year 2006, 135,551 people lived in Trinidad (373 were interviewed). There
were 55.27% men and 44.73% women. Inhabitants were grouped in 32,103 households,
with an average number of 5.3 members. According to language spoken, nobody living in
Trinidad is, or might be considered indigenous. According to self-identification
parameters, 25.09% from total population might consider themselves belonging to some
indigenous or native people. In year 2007, 131,140 people lived in Trinidad (333 people
were interviewed); 50.75% men and 49.25% women.
Poverty. Following are displayed poverty levels according to poverty rates according to
Foster-Greer-Thorbecke (FGT(a)) poverty rates.
Chart 4.3.
INCIDENCE, GAP AND INTENSITY
POVERTY INDEXES IN THE CITY OF TRINIDAD
Año
2001
2005
2006
2007
a=0
0.54906
0.35121
0.3775
0.4932
a=1
0.18757
0.11862
0.15133
0.21601
a=2
0.08512
0.04898
0.07315
0.12455
Where: FGT(0) = headcount ratio (proportion of poor)
FGT(1) = average normalised poverty gap
FGT(2) = average squared normalised poverty gap
According to previous chart, poverty levels between years 2006 and 2007 increased in
almost 12 points, accounting for large shock on people of Trinidad. Similar situation may
be observed in the case of poverty gap, with inter-annual increase above 6 points. Finally,
poverty severity, expressed by FGT(2) index has also undergone an increase between years
assessed.
28
These results reveal enormous affectation in Trinidad living conditions, supposedly derived
from natural disaster effects. Increase of poverty incidence in almost six times above
Bolivian average may be observed, when comparing these figures with national average.8
Therefore attempts should be made to separate effect of inundation on population living
conditions from assessment of household income.9
Income Patterns
Following methodology described in section 2, first step should be estimating income
regressions, taking the log of income per hour per person coming from their working
activity during month previous to survey as dependent variable
Available independent variables for both years are10:

EDUCA individual highest grade of finished studies

Exp and exp2 which is a proxy of potential working years experience of a person,
calculated as age minus number of years of finished studies, minus six. Linear as
well as a quadratic form are included.

Sex, a dichotomic variable identifying person sex.

District. Which is equal to one, if household lives within city’s first protection ring
against inundation (artificial barrier constructed to prevent water advance) and zero
in any other case.11 This is the variable that allow to capture, at least partially, the
effects of inundation on individual well-being.
Results of regression are displayed in Chart 4.4. As is has been explained, estimations do
not represent individual follow up in time, since this is not panel data. However, results are
consistent.
8
FGT(0) increase for the whole of Bolivia was only two points in the reference period.
According to theory (Borjas,1997), the district of residence is an endogenous variable to the household income, because
we are in presence of double causality problem. On the one hand, income determines the place of residence of a
household, according money possibilities to choose a better or worse location. On the other hand, there are effects of
district residence determining household outcome in education and income achievement. Though, this case is a cuasiexperiment situation deriving from the fact that we are in presence of exogenous shock. Accordingly to this, it is expected
that, explaining power of this variable may mainly come from event affectation on household living conditions,.
10
Language variable is not included since household survey does not report indigenous tongue. Besides, self identification
question report that only 1.5% of surveyed declared themselves belonging to an indigenous group.
11
Census zones within protection ring against inundations are 9, while remaining 8 are outside and were most affected by
2006-2007 inundation.
9
29
CHART 4.4.
INCOME FUNCTIONS
Dependent Variable: Labour Income Logarithm
2006
2007
Education
Sex
Experience
experience Sqr.
District
Constant
R2 adjusted
F
Obs.
.0879574
(.0120931)
-.3877122
(.0959541)
.0440748
(.0078642)
-.0004251
(.0001388)
.0500732
(.01407258)
5.925533
(.2244203)
.0667194
(.0171117)
-.6491378
(.1165754)
.040709
( .011347)
-.0004794
(.0002152)
.0719007
(.01196048)
6.68306
(.2877115)
0.3255
18.47
182
0.2614
10.63
137
Regression results displayed in Chart 4.4 indicate that for every year of schooling, there is
an average income increase between 6.6% and 8.7%. It should be noted that return to
school was significantly lower in year 2007. This situation might be due to the fact that
disaster affected city productivity negatively, and consequently also affected return of
education on income generation.12
Similar situation may be found in the case of sex variable. In fact, it is observed that
income earned by men is above that of women’s in those two years. Additionally, this
negative effect on women increased between 2006 and 2007, implying that changes in the
city’s conditions along that period – including inundation event – have had a clearer effect
on feminine population.
With regards to working experience, we can note that increases in experience have a
positive return on income at decreasing rate. Changes in both years are not statistically
different for the case of Trinidad.
Finally, it may be observed that the fact of living in a district protected from inundation
(least exposed to risk) has a positive income return. Although, nothing may be concluded
about the mechanisms that produce such extraordinary return, it is evident that this
coefficient grew from 5% to 7% between 2006 and 2007.
12
Results for this and other variables are valid as long as regression and decomposition are specified in terms of monthly
incomes instead of hourly income, or whenever depending variable is expressed in terms of levels, instead of logarithms.
30
Income Inequality and Decomposition
Once returns from several income determinants have been assessed, identifying their
meaning in terms of functionality is of interest. It has to do with decomposing income
inequalities to determine main factors differentiating high income, from low income
workers. To this effect, Fields decomposition discussed on section 2 was used.
First step for this type of analysis was Gini (GE(1)) coefficient calculation for both years.
In the reference period, this inequality measure varies between 0.37 and 0.43, situating
Trinidad within an intermediate range in terms of inequality, with respect to national
average, as well as in comparison to all other capital cities. 13 Nonetheless, the
extraordinary growth of this index – almost five points – induce to consider that natural
event registered in this city, did not only lead to an increase in poverty incidence, but it
also affected population in term of inequality. It is presumed that most vulnerable workers
to event were those that got the greater inundation impact.
Results from decomposition are shown in Chart 4.5
CHART 4.5.
FACTOR’S CONTRIBUTION TO LABOR INCOME INEQUALITY IN TRINIDAD
GINI




Education
Experience
Sex
District
2006
2007
0.3781
0.4381
Variable
0.774
0.079
0.073
0.074
0.755
0.067
0.079
0.096
Decomposition exercise shows that schooling (education) is the most important variable in
order to explain income inequality in Trinidad. Schooling accounts for between 75% and
77% of inequality. This factor, that is over 9 times greater than other factors, including
district of residence. All other variables together explain only one fraction of what has been
explained by education.
Nevertheless, registered changes in all other factors between 2006 and 2007 should be
noted. In the first place, second important factor in inequality decomposition for year 2006
is working experience, accounting for 7.9 % of inequality. District residence and sex take
third and fourth places this year, with 7.4% and 7.3% of contribution to inequality
respectively. The magnitude of factors contribution for 2007 registered an important
change. Though education still being the main factor in order to explain inequality in that
year, district factor takes the second place now, with a 9.6% of contribution to inequality.
13
This is consistent with previous surveys: Urquiola (1993) shows Gini coefficients around 0.5 % for urban areas in
Bolivia and World Bank (2005) and Hernani (2005) believe this value nears 60% for all of Bolivia.
31
Both education as well as experience decrease their contribution to inequality, and only sex
displays an increase, which is though, less significant than district residence.
These results allow inferring that district residence contribution to inequality could be
partly due to effect of natural disaster impact on the city of Trinidad. Additionally,
increased sex contribution reveals that women’s conditions are not only worse than men’s
in absolute terms, (lower return) but also that inequality affects women more than men after
natural events.
32
5. Conclusions
The objective of this document has been to clarify relation between Bolivian well-being
measures and natural disaster risks. However, limited data lead to circumscribe the analysis
to relatively general conclusions and estimations.
Municipal analysis of the city of Trinidad does not permit answering all questions related
neither how to attenuate impact on individual well-being exposed to disaster risk. Still,
information and analysis indicate that effect on income and inequality is important.
From results obtained by the present work, some important conclusions may be inferred.
Municipal changes on population well-being are negatively correlated to natural disaster
events. Though causal relations may not be establish between natural events and change in
poverty levels, it is quite clear that municipalities more exposed to natural disaster are most
prone to worsen off well-being conditions. The same conclusion can be observed from
results of typologies of municipalities under natural events exposition criteria, that is, those
municipalities most affected by intense events (even without having data on impacts and
damages) are the most affected in their well-being.
In Bolivia, analysis on natural disaster tendencies shows us that over 70% of the country’s
population is exposed to some type of event. The most important disasters that affect
population are freezing, drought and inundation. The Altiplano is Bolivia’s most affected
region, though it is not one undergoing greatest impact. Amazone region, instead, though
not being most affected in number of events, is the one undergoing greatest impact, on
account of events extent.
Assessment of poverty indicators at 1992-2001 inter-census periods reveals that
municipality hazard incidence is correlated to increases on poverty. Although is not
possible to isolate the effects; estimations show that poverty increased about 3% (at the
most conservative estimations) in regions exposed to natural events. Statistical data also
shows that hazards do not have a negative effect on economic well-being indicators alone,
but also on non economic ones (UBN). These conclusions on well-being indicators should
be considered by policy making officials, since issues like schooling assistance, housing
improvement indexes and access to basic services – the main components of UBN –
undergo as strong or stronger impact than economic matters like productive infrastructure
and physical capital.
With regards to case study on the city of Trinidad, it may be established that change in
population well-being conditions, was so deep after 2006-2007 inundation that it is unlikely
that an additional determinant might be found besides magnitude of disaster. In effect,
changes in well-being conditions of the city of Trinidad are statistically different to those
observed in urban zones of the rest of the country. Even more important is the fact that
poverty incidence in this period grew in 12 points in the city of Trinidad, notably higher
than national average.
33
It is not only that poverty levels display important increase, but it is also that income
distribution statistics rise significantly. Gini index for labour income in Trinidad rose from
0.38 % to 0.43 % during the mentioned period, showing a presumable outstanding impact
of natural event on lower income population.
In term of inequality, it is evident that although education is the main component to explain
inequality, exposition to risk - approached in this document through Trinidad district of
residence – are relatively important, once natural events have occurred.
Assessment of relations between well-being levels and disaster hazards reveals that there
are still many questions to be answered: which issue of well-being conditions should be
first approached for hazard prevention and recuperation? How to choose suitable insurance
mechanism of disaster hazard in order to minimize impact? Which should be the most
important action after emergency situation? Careful evaluation of development and
emergency response policies should be face at the light of indicative results displayed in
this paper.
Finally, limited analysis in the case of Bolivia, shows the need for improved information,
both on natural events and hazards as well as on aspects related to disaggregation of wellbeing data.
34
References
• Borjas, G. (1997), “To Ghetto or Not to Ghetto: Ethnicity and Residential Segregation”,
NBER Working Paper 6176. Cambridge, United States: National Bureau of Economic
Research.
• CEPAL (2007), Alteraciones Climáticas en Bolivia, Impactos observados en el primer
trimestre de 2007, La Paz, Bolivia.
• CEPAL (Buro de Prevención de Crisis y Reconstrucción / Programa de Naciones Unidas
para el Desarrollo) (2006), Marco Estratégico para la Planificación de la Recuperación y la
Transición al Desarrollo-Inundaciones y Granizadas en Bolivia 2006, Programas generales
de intervención y presupuesto, La Paz, Bolivia.
• De la Fuente, Alejandro et al. (2008), Assessing the Relationship between Natural
Hazards and Poverty: A Conceptual and Methodological Proposal, PNUD-ISRD, Mimeo.
• Fields, Gary S. (1997), Accounting for Income Inequality and Its Change, Paper presented
at the annual meetings of the American Economic Association, New
Orleans, January.
• Fields, Gary et. Al (1997), Descomposición de la desigualdad del ingreso laboral en las
zonas urbanas de Bolivia, UDAPSO, La Paz, Bolivia
• Hernani, Werner, (2005), Mercado Laboral, Pobreza y desigualdad en Bolivia, INE, La
Paz, Bolivia.
• Litchfield Julie (1999), Inequality, Methods and Tools, World Bank, Washington D.C.
• Ministerio de Desarrollo Sostenible, (1997), Sistema Nacional de Información de
Ordenamiento Territorial, La Paz, Bolivia.
• PMA, (Programa Mundial de Alimentos), (2008), Serie Creación de capacidades para
respuestas a emergencias. Plan departamental de contingencias ante la crisis alimentaria por
emergencias. SEQUIA-INUNDACION. Beni. PMA-Defensa Civil.
• PNUD (Programa de la Naciones Unidas para el Desarrollo) (2004), Dirección de
Prevención de Crisis y de Recuperación, La reducción de Riesgos de Desastres. Un
Desafío para el Desarrollo. Un informe Mundial, [en línea]
http://www.undp.org/bcpr/disred/rde.htm.
35
• PNUD, (2007), Objetivos de Desarrollo del Milenio. Beni: Situación antes del
Fenómeno de El Nino, La Paz, Bolivia.
• UDAPE (Unidad de Análisis de Políticas Sociales y Económicas) (1993), Mapa de la
Pobreza en Bolivia: un Instrumentos para la acción, La Paz, Bolivia.
• UDAPE (2002), Mapa de la Pobreza en Bolivia – Necesidades Básicas Insatisfechas
2001, La Paz, Bolivia.
• UDAPE (2006), Pobreza y desigualdad en los municipios de Bolivia. Estimación del
gasto de consumo combinando el Censo 2001 y las encuestas de hogares. 3a ed. La
Paz: Instituto Nacional de Estadística.
• Urquiola, Miguel (1993), Aproximación a los determinantes de la distribución personal
del ingreso en el área urbana de Bolivia, Documento de Investigación, UDAPSO, La Paz,
Bolivia.
• World Bank (2005), Bolivia Poverty Assessment, Report No. 28068-BO, Washington
D.C.
36
ANNEX 1. WELL-BEING MEASURES
Following several well-being measures were undertaken – both at municipal as well as
at household levels – based on available official information:
Dimension
Income
Indicator
Poverty incidence (poverty
line)
Poverty gap
Inequality
Basic Needs
Unmet Basic Needs Index
Human
Development
Development Index
Malnutrition
Global Malnutrition Rates
Definition
Households
Unable to
have Access to
Basic Set of
Good “Typical
Consumption
Set” according
to Family
Budgets and
Family
Surveys
Household
Distance from
Poverty Line.
Household
Income /
Consumption
Distribution.
Poverty
Indicator
Constructed
by Normative
Criteria on
Education,
Health,
Housing
Quality and
Access to
Basic
Services.
Composed
Index
Assessment
According to
Life
Expectancy,
Per Capita
Income, and
Education
Issues
(Adequacy
and Illiteracy)
Variables.
Weight /
Source
Income Variable obtained from
INE Household Surveys and
1992, 2001 CENSUSES Income
Imputations.
Income Variable obtained from
INE Household Surveys and
Income Imputations on 1992
and 2001 CENSUSES.
Income Variable Obtained from
INE Household Surveys and
Income Imputations on 1992
and 2001 CENSUSES.
1992 and 2001 Population
CENSUSES.
Consumption and Income
Imputation from National
Accounts, CENSUSES Data,
Administrative Registers from
the Ministry of Education and
from Household Surveys on
Education Issues. Demography
and Health Surveys towards Life
Expectancy Estimations.
CENSUSES and Health Surveys
37
Life
Expectancy
Life Expectancy at Birth
Indicator
Height
Relation
Among
Children
Between 2 and
5 Years When
Compared to
Normative
Values.
Life
Expectancy
Elaborated
from Living
Normative
Tables
Index Imputations
Population CENSUSES.
38
ANNEX 2: DESCRIPTION OF PRODUCTIVE AND CLIMATIC VARIABLES
1. Agricultural Potential
Agricultural potential indicator refers to soil aptitude for agricultural activities
development. It may be divided into four categories, according to the following Chart:
Agricultural Potential Categories
Meaning
Categories
Optimal (unlimited)
1
Moderate (moderate limitations)
2
Very low (severe limitations)
3
Limited (very severe limitations)
4
Source: Planning and Sustainable Development Ministry, 1997.
PMA information was used for VAM 2002 assessment based on Agricultural Potential Map
by the Planning and Sustainable Development Ministry on 1997. This map was digitalized
and superposed on map incorporating municipal political and administrative division.
2. Forestry Potential
This indicator refers to determining apt soil for forestry activities, by means of quantifying
number of cubic meters of forestry produce per hectare and per year. It is divided into the
following categories:
Meaning
Poor Potential
Low Potential
Limited Potential
Medium Potential
High Potential
1
2
3
4
5
Forestry Potential Categories
Categories
M3 Hectare Produce
1-5
5-7
7-9
9-11
11-14
Source: Ministry of Planning and Sustainable Development, 1997.
Data source and assessment procedures are similar to those of Agricultural Potential.
3. Geographic and Climatic Characteristics
Indicators detailed in following Chart were obtained from INESAD data base:
39
Geographic and Climatic Characteristics
Indicator
Definition
Altitude
It Refers to Average Meters Above Sea Level a Municipality is
Located. It is Expected that to Greater Altitude, Increased
Vulnerability to Alimentary Insecurity Follows
It is the Quantity and Amount of Municipal Rainfall Measured in
Centimetres Per Year. It is Expected that More Frequent Rains,
While These Do Not Turn into Inundation, will be Help Reduce
Alimentary Insecurity Vulnerability.
Fluvial Precipitation
Road Density
It refers to the Amount of Kilometres of Main Roads in
Territorial Area. It is Expected that Municipalities with Better
Connections to Roads be Least Vulnerable to Alimentary
Insecurity.
Following indicators were obtained from same source as Agricultural and Forestry
Potential. Classification and Categories are detailed as follows:
Climatic Indicators
Indicator
Category
Meaning
Drought Frequency in Years
Low
Medium
High
Very High
Very Low
Low
Medium
High
1 Every 10 Years
1 Every 5 Years
1 Every 2 Years
4 Every 5 Years
No freezing
30 to 90 days Freezing per Year
90 to 180 days Freezing per Year
180 to 270 days Freezing per
Year
270 to 330 days Freezing per
Year
No Inundations Problems
Less
than
30%
Inundated Surface
Between 30% to 50%
Inundated Surface
Freezing Days per Year
Very High
Area under Inundation Hazards
Low
Medium
High
Very High
Over 50% Surface
40
ANNEX 3: INEQUALITY DECOMPOSITION METHODOLOGY
Fields 1997 and Fields et. Al (1997) proposes a method for decomposing sources of income
inequality. His method is based on an income generation function and concludes with
following answers: “x% inequality income is explained by education and % by area
residence, z% by gender of the individual, etc”.14
From an income generation function, based on human capital theory, or some other
theoretical model sustaining it, in which i logarithm of an individual’s income in period t is
specified as a function of a series of explaining variables (identified by sub-index j),
ln(Yit)= t + jjtwijt + t
(1)
which may be rewritten as ln(Yit)= jajtzijt = a’Z
Where a = ( 1 2... J 1)
Z = (1 x1 x2...xj )
(2a)
(2b) and
(2c)
Strategy for obtaining a useful decomposition equation consists in decomposing an
inequality measure, income log-variation in this case, to further explain that same
decomposition may be also applied to other inequality measures.15
By using inequality functions in (2a) and (2c), variant is taken from both sides. To the
left income log-variant is obtained, while right side variant may be manipulated to obtain
following result:
Result N1. Since income generation function (2ª – 2c), is sj (In Y) and proportion of
income log-variant attributed to explaining factorj, covariant will be called cov(.) and
variant o2(.) Income log-variable may then be decomposed as:
sj(ln Y)= cov(ajzj,ln Y)/ 2(ln Y)
(3a)
where,
(3b) and
(j=1,J+1) sj(ln Y)= 100%
(j=1,J) cov(ajzj,ln Y)/ 2(ln Y)= R2(ln Y)
(3c)
If pj (In Y) is fraction of log-variable applied to j explaining factor, it is concluded that
pj(ln Y) sj(ln Y)/ R2(ln Y)
(3d)
14
Income Generation Function Term is used instead of Income Functions or Salary Equations owing to Methodology
being Sufficiently Ample for Non Labour Income to be Included together with Labour Income in Regression, if
Researcher Considers it Appropriate.
15
Log Variant is Income Logarithm Variant, Not Variant of Income Logarithm, as is sometimes wrongly interpreted.
41
Furthermore, other inequality measures other than log-variant may be decomposed. To this
purpose, result belonging to inequality decomposition literature through factorial
components may be used. On this literature, total income Yi, from i-ism receptor unit is
represented as income addition resulting from each different factorial component, i.e.,
labour income, INCOME PER CAPITA, income from transferences, etc. This is as follows:
Yi = kYik
(4),
Then, it must be determined which fraction from total inequality, represented by an
inequality measure (Y1….Yn), may be explained by labour produced income, capital
income, income from transferences, etc.
Sk Relative contribution to factorial inequality is established, as income inequality
percentage explained by k-ism factor, u is Ei(Yi/n), average income. )Shorrocks, 1982) an
important theorem on factorial component decomposition, displays following:
Result N 2. Under six axioms enumerated in appendix, sk, relative contributions to
factorial inequality are given by:
sk = cov(Yk,Y)/2(Y)
de manera que
 k sk = 1
for any l(Y1….Yn) inequality index defined on income vector (Y1…Yn) as long as index
is continuous and symmetric, and satisfies condition l( u,u----,u) = 0.
Above decomposition may be applied to almost all inequality measures, including Gini
coefficient, Atkinson index, generalized entropy indexes family, log-variant and several
measures based on percentiles.
Shorrocks theorem may now used to decompose labour income inequality, from income
generation functions. These functions are:
ln(Yit)= jajtzijt = a’Z
(2a)
while function expressing total income as income addition of each component is,
Yi = kYik
(4);
both are simple additions. It must also be noted that on decomposing inequality of (4),
Schorrocks obtains
sk = cov(Yk,Y)/2(Y), así como k sk = 1,
42
which is the same as (3) with Yk replacing ajzj and with Y instead of In(Y). Through this
homeomorphism and applying Shorrocks theorem following result is achieved:
Result No. 3. Given (2ª 2c) income generation , I)InY) inequality index is established on
income logarithm vector, InY=(InY1,…,InYn).
Under six enumerated axioms in appendix, income inequality decomposition given by:
sj(lnY) = cov(ajzj,lnY)/2(lnY)
where
(3a)
(j=1,J+1) sj(ln Y)= 100%
(3b)
(j=1,J) cov(ajzj,ln Y)/  (ln Y)= R (ln Y)
2
2
2
And where pj(lnY) sj(lnY)/R (lnY)
(3c)
(3d)
is maintained not for log-variant alone, but for almost any (In Y1,…,InYn) inequality index
that is continuous and symmetric, and as long as long as it meets condition that
l(u,u…,u)=0.
Measures subject to this decomposition include Gini-log and Atkinson-log, generalized
entropy log-family and log-percentiles.16
Result 3 states that if decomposition conditions are accepted and if inequality measure
decomposition based on income logarithm vector is also accepted, then it is not compulsory
to restrict oneself to a specific inequality measure for decomposition. This is due to the fact
that all UNDERLINE ALL inequality measures that might be considered useful would have
same j ism explaining factors percentage effects, when measure is applied to income
logarithms.
In order to prevent violation of transference principle expressed in the usual way, human
capital theory may be abandoned to use income levels instead of logarithms for the function
of income generation. When this is the case, regression coefficients as well as percentage
contributions will change. Following result is obtained when using b to denote this
proceeding’s coefficient:
Result N. 4. Given income generation
Yit= jbjtzijt = b’Z
Where b = ( 1 2... J 1)
Z = (1 x1 x2...xj )
(2a’)
(2b’) and
(2c’),
L(Y) inequality index is determined on Y =(y1….yn) inequality vector on six enumerated
axioms in appendix, income inequality decomposition given
16
In all cases, where “log-l” Measures are mentioned, reference is made to Income Logarithms.
43
sj(Y) = cov(bjzj,Y)/2(Y)
(3a’)
en donde (j=1,J+1) sj(Y)= 100%
(3b’),
(j=1,J) cov(bjzj,Y)/  (Y)= R (Y)
(3c’)
(3d’)
2
2
2
y en donde pj(Y) sj(Y)/R (Y)
is maintained for any continuous and symmetric inequality index l(Y1,…Yn), as long as
l(u,u…,u)=0.
Measures that may be decomposed this way include Gini coefficient, Atkinson index,
generalized entropy index family and measures based on percentiles.
Results 3 and 4 together explain that: i) log-variance, Gini log, Theil log, etc., all provide
same j-ism explaining factor percentage contributions factor to income inequality
logarithm, ii) Gini coefficient, Theil index, etc variances, at their ordinary form, provide
same percentage contributions to j-ism explaining factor to income inequality, however, iii)
– percentage contributions – answers in i) and ii) are not the same.
44
ANNEX 4: UNMET BASIC NEEDS ESTIMATION
Unmet Basic Needs Index (NBI by its acronym in spanish) is assessed through lack of
various needs. These needs are: education, access to health services, housing materials,
overcrowding, and basic draining and energy services.
Lack Indexes. (NBI) lack indexes were assessed according to NBI and INE qualification
methodology. According to same
nbi( x) 
Where:
Nbi(x)
Nx
Cx
Nx  Cx
Nx
component lack index
component normative value
household x observed component
NBI (x) might be rewritten as follows:
nbi( x)  1 
Cx
Nx
For practical purposes Lx will be called Observed Goal Coefficient divided by Norm will
be called )Lx) Achievement Goal:
Lx 
Cx
Nx
Then
Nbi(x) =1 - Lx
Lack Index displays dissatisfaction level or degree with regards to normative values.
By definition it may take a value from (-1, 1) range, where:
Lack Indicator
Levels of satisfaction
-1
levles of dissatisfaction
0
1
Minimal level of satisfaction
Positive values denote dissatisfaction levels while those nearing unity indicate greater lack,
on the contrary, negative values reflect over minimal satisfaction level, and the more they
tend to -1, greater satisfaction will be met.
45